> '` bjbj$$ .FF~ v
v
v
v
QQQ
$h,x
!_QQ__
v
v
)
)
)
_:v
v
)
_
)
)
)
v
>a :)
0)
@)
)
XQ0")
QQQ
QQQ____$Temple University
Department of Economics
Economics 8190
Propensity Score and Matching Homework
1. The data in JTRAIN2 are from a job training experiment. The goal is to determine the average treatment effect from participation in the training program. The notation that we will use is: The variable "earn78" are real earnings in 1978, measured in thousands of dollars. The variable w is an indicator to show participation in the training program. The vector X includes age (A) and its square, real earnings (earn) in 1974 and 1975, a high school graduation dummy (HS), a dummy for marital status (M), and dummies for black (B) and Hispanic (His).
a. Compute the average of earn78 for those who get training and for those who do not get training. Is the difference between the means different from zero?
b. Repeat question a., but do the training and not training means for HS grad (or not), marital status, and race.
c.1. Estimate the propensity function as a linear probability model and save the fitted values of w, we'll call them EMBED Equation.3 .
EMBED Equation.3
c.2. Estimate propensity functions using the same RHS variables, but use logit and probit. Save the fitted values for w in both cases, we'll call them EMBED Equation.3 and EMBED Equation.3 .
d. Estimate the unknown coefficients of the regression:
EMBED Equation.3
e. Using the fitted propensities estimate the unkowns in the following two equations:
EMBED Equation.3
EMBED Equation.3
You will have six sets of coefficient estimates.
f. Compare and contrast the different results that you got for the treatment effect across the many models of this exercise.
2. In class we reviewed the paper by Buck, Friedman and Dunkelberg about small business loans. In that paper there was a variable on the RHS that was a dummy variable for whether or not the firm's last loan application was accepted. Since that variable was likely to be correlated with the error term it was replaced by Heckman's . How does this approach differ in substance and interpretation from the Average Treatment Effect approach that you used in question 1?
;f $ 6 8
:
?
[
_
v
z
-.+,-.12EFGHI߫ߜۀjrPP
hzCJUVaJjh.* h.* EHUjoPP
h.* CJUVaJjh.* UjhzhzEHUjrPP
hzCJUVaJjhzUh7hzh.* hT^vH*h'hT^vh!(hl=hihh[1*+:;bc
./01I
P
Q
j
k
%
R
S
f
g
h
i
滮槣{s{dWs{ShnDtjh!(h!(EHUj>PP
hT^vCJUVaJjhVbUhVbhT^vhzj; h.* h(tEHUjSsPP
h(tCJUVaJj'hzhzEHUjrPP
hzCJUVaJhzjhzUjhzhzEHU %&,1h/ =!"#$%Dd
Tb
c$A??3"`?2^صAP.
R:D
`!2صAP.
R HXJxcdd`` @c112BYL%bpu1,/MLDd
tb
c$A??3"`?2IίXS ˒<4!$%X
`!ίXS ˒<4!$0 /
Ixcdd```b``baV d,FYzP1n:L
B@?b 10sC0&dT20$ KX2sl/"6~.Ob3D^t9M o%'?[,NXR /y#COdfjbQLdLg57D'B|.s#w0Oug' U y80X7M/Fp\l !70L@(\\bXvbkp:<2<<,R1T
QTyE~T/\8PP9YQEQos 4oq8As;c+KRs^CDl,Ā rDd
@Tb
c$A??3"`?2^6s3͋&s:W
`!26s3͋&s XJxcdd`` @c112BYL%bpukʄٕ7nؙ>Ƴ3Ʋ:piUjƞ{}E6UsI[[rVLPYxw]<;_e+~!i~"jANhj%/~Foj>mT|[~a] ǵ|C9Zܯ9?@:<$^q^-8rO1PRpbM/Sz_5?>y ۚ'*)-npN!Oܞ@o3[2L:h}Yb.YgbDd
hb
c$A??3"`?2~
,RNCt
`!~
,RNCt@R|lxcdd``cd``baV d,FYzP1n:&B@?b ~
ㆪaM,,He`H @201(8e&٫",>
1Dl\
_dXr@@ڈHf%ď-|
8?/p.7a٘Xq.bF9.da"o̠6RL@(\H|!b
`{,` 6 'AlYA|[.h~~aa_1347449510F>a>aOle
CompObjfObjInfo
!"#&)*-012456789;<=>@
FMicrosoft Equation 3.0DS EquationEquation.39q8D
2p 1
FMicrosoft Equation 3.0DS EquationEquation.39qEquation Native :_1347448809F>a>aOle
CompObj
fObjInfo
Equation Native e_1347449556F>a>aOle
I8D
w=0
+1
A+2
A2
+3
earn74
+4
earn75
+5
HS+6
B+7
His+8
MCompObjfObjInfoEquation Native :_1347449565F>a>a
FMicrosoft Equation 3.0DS EquationEquation.39ql
2p 2
FMicrosoft Equation 3.0DS EquationEquation.39qOle
CompObjfObjInfoEquation Native :
2p 3
FMicrosoft Equation 3.0DS EquationEquation.39qs8D
earn78
=0
+w+1
A+2
A2
+3
_1347449683F>a>aOle
CompObjfObjInfoEquation Native _1347436077" F>a>aOle
$CompObj %fearn74
+4
earn75
+5
HS+6
B+7
His+8
M
FMicrosoft Equation 3.0DS EquationEquation.39qObjInfo!'Equation Native (_1347435582$F>a>aOle
+w(<2
earn78
=0
+w+1
2p (X)
FMicrosoft Equation 3.0DS EquationEquation.39q=
earn78CompObj#%,fObjInfo&.Equation Native /1Table-
=0
+1
w+1
2p (X)+2
w(2p (x)"" 2p
)Oh+'0
<H
T`hpxTemple UniversityAndrew BuckNormalAndrew Buck`Tvl?@A.n:H'uTI'քK/;^3
x5uHVU.Mvc2YN1tV UF9kgm<B$_6|Pw/>
[g\-nYw\m7B>u0 rexA~:\&4벿GMrZux+Y`a~q &*J^xrV+{p}11g"G)t16>T|mDLW'&;>"8ߎl 5_".i+/f?ڄ~~